CN117407817A - An abnormality monitoring system for distribution automation machine room - Google Patents
An abnormality monitoring system for distribution automation machine room Download PDFInfo
- Publication number
- CN117407817A CN117407817A CN202311398096.7A CN202311398096A CN117407817A CN 117407817 A CN117407817 A CN 117407817A CN 202311398096 A CN202311398096 A CN 202311398096A CN 117407817 A CN117407817 A CN 117407817A
- Authority
- CN
- China
- Prior art keywords
- model
- time sequence
- machine room
- model parameter
- data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/2433—Single-class perspective, e.g. one-against-all classification; Novelty detection; Outlier detection
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/254—Fusion techniques of classification results, e.g. of results related to same input data
- G06F18/256—Fusion techniques of classification results, e.g. of results related to same input data of results relating to different input data, e.g. multimodal recognition
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Business, Economics & Management (AREA)
- Health & Medical Sciences (AREA)
- Economics (AREA)
- General Health & Medical Sciences (AREA)
- Public Health (AREA)
- General Business, Economics & Management (AREA)
- Water Supply & Treatment (AREA)
- Tourism & Hospitality (AREA)
- Strategic Management (AREA)
- Primary Health Care (AREA)
- Marketing (AREA)
- Human Resources & Organizations (AREA)
- Testing And Monitoring For Control Systems (AREA)
Abstract
The invention discloses an abnormality monitoring system of a power distribution automation machine room, which relates to the technical field of intelligent monitoring, and can realize the collection and intelligent analysis of multidimensional sensing data in the power distribution automation machine room, so that the intelligent monitoring of the machine room can be realized, the result of the intelligent analysis can be remotely shared to staff, the staff can timely grasp the parameters of the machine room operation, the abnormal condition of the machine room operation can be timely found, and various abnormal conditions can be rapidly processed, thereby effectively reducing the property loss caused by the abnormal condition of the power distribution automation machine room.
Description
Technical Field
The invention relates to the technical field of intelligent monitoring, in particular to an abnormality monitoring system of a power distribution automation machine room.
Background
The power distribution automation machine room is a core part of a power distribution automation system, and if the power supply of the machine room loses voltage and the temperature is too high, equipment such as a server of the power distribution automation system and the like can be stopped, the system is completely paralyzed, and the consequence is not considered. In order to ensure the safe operation, workers are required to patrol the machine room, wherein the alternating-current power supply voltage and the machine room temperature are the most important two patrol items of the power distribution automation machine room, but manual patrol is carried out periodically, the real-time remote monitoring of the alternating-current power supply voltage and the machine room temperature of the machine room cannot be carried out, and the existing potential safety hazards cannot be found timely.
The current machine room management device is a dynamic ring system, and the monitoring object of the dynamic ring system is mainly machine room power and environment equipment (such as power distribution, air conditioning, temperature and humidity, water leakage, entrance guard, security protection, fire protection and the like). However, the existing monitoring system can only monitor the running state in the corresponding management system, cannot monitor the whole running environment of the machine room, and cannot push alarm information to management staff in time when faults occur, so that serious potential safety hazards exist in the machine room, and larger property loss is easy to cause. In addition, the existing power distribution automation monitoring system does not have reference indexes for providing daily overall operation conditions of the machine room, so that a manager cannot timely master the operation change conditions of the machine room.
On the other hand, the traditional fire alarm controller at present only sends an audible and visual signal on site, which is unfavorable for people to quickly know that a fire condition occurs, and is unfavorable for people to fight for time to extinguish and escape.
Disclosure of Invention
The invention aims to provide an abnormality monitoring system of a power distribution automation machine room, which solves the problems existing in the prior art.
The invention is realized by the following technical scheme:
an abnormality monitoring system of a power distribution automation machine room comprises a data sensing module, a model acquisition module, a data analysis module, an online data storage module and a remote inquiry module;
The data sensing module is used for collecting multidimensional sensing data of the power distribution automation machine room and transmitting the multidimensional sensing data to the data analysis module;
the model acquisition module is used for constructing an abnormal recognition model by adopting multi-model combination, training the abnormal recognition model by adopting a multi-stage reinforcement training algorithm, acquiring a trained abnormal recognition model, and deploying the trained abnormal recognition model in the data analysis module;
the data analysis module is used for scheduling a pre-deployed trained anomaly identification model to analyze multidimensional transmission data, acquiring anomaly monitoring results of power distribution automation and a method, and transmitting the anomaly monitoring results to the online data storage module for storage;
the online data storage module is used for receiving and storing the abnormal monitoring results transmitted by the data analysis module and sharing the abnormal monitoring results to the remote inquiry module in an online checking mode;
the remote inquiry module is deployed on a mobile terminal of a worker and is used for remotely accessing the abnormal monitoring result stored in the online data storage module.
In one possible implementation, the multi-dimensional sensing data includes a machine room temperature time sequence, a machine room power supply voltage time sequence, a water level sensing time sequence, a smoke sensing time sequence, and an image sensing time sequence, which are acquired at a preset sampling frequency.
In one possible implementation manner, the model acquisition module comprises a model generation unit, a training data acquisition unit, a model training unit and a model transmission unit;
the model generation unit is used for generating sub-models respectively corresponding to the machine room temperature time sequence, the machine room power supply voltage time sequence, the water level sensing time sequence, the smoke sensing time sequence and the image sensing time sequence, and constructing an abnormal recognition model according to all the sub-models;
the training data acquisition unit is used for acquiring training data input by man-machine interaction or pre-stored in a database;
the model training unit is used for training the abnormal recognition model by adopting a multi-stage reinforcement training algorithm according to the training data acquired by the training data acquisition unit to acquire a trained abnormal recognition model;
the model transmission unit is used for deploying the trained abnormal recognition model in the data analysis module.
In one possible implementation manner, before generating the sub-models respectively corresponding to the machine room temperature time sequence, the machine room power supply voltage time sequence, the water level sensing time sequence, the smoke sensing time sequence and the image sensing time sequence, and constructing the abnormality recognition model according to all the sub-models, the method further comprises:
Determining whether the temperature data exceeds a preset temperature threshold value according to each temperature data in the machine room temperature time sequence, if so, directly determining temperature abnormality to obtain a temperature abnormality monitoring result, otherwise, constructing a submodel;
for each voltage data in the machine room power supply voltage time sequence, determining whether the voltage data exceeds a preset voltage threshold, if so, directly determining voltage abnormality to obtain a voltage abnormality monitoring result, otherwise, constructing a sub-model;
for each water level data in the water level sensing time sequence, determining whether the water level data exceeds a preset water level threshold, if so, directly determining water level abnormality to obtain a water level abnormality monitoring result, otherwise, constructing a submodel;
for each smoke data in the smoke sensing time sequence, determining whether the smoke data exceeds a preset smoke threshold, if so, directly determining smoke abnormality to obtain a smoke abnormality monitoring result, otherwise, constructing a submodel;
and determining a first abnormal monitoring result according to the temperature abnormal monitoring result, the voltage abnormal monitoring result, the water level abnormal monitoring result and the smoke abnormal monitoring result.
In one possible implementation manner, generating sub-models respectively corresponding to the machine room temperature time sequence, the machine room power supply voltage time sequence, the water level sensing time sequence, the smoke sensing time sequence and the image sensing time sequence, and constructing an anomaly identification model according to all the sub-models, including:
generating a first feature extraction sub-model for extracting features in a machine room temperature time sequence, a machine room power supply voltage time sequence, a water level sensing time sequence and a smoke sensing time sequence, and generating a second feature extraction sub-model for extracting features in an image sensing time sequence;
generating a feature recognition sub-model for performing fusion recognition on the features output by the first feature extraction sub-model and the second feature extraction sub-model;
and constructing an anomaly recognition model according to the first feature extraction sub-model, the second feature extraction sub-model and the feature recognition sub-model.
In one possible implementation manner, training the anomaly identification model by using a multi-stage reinforcement training algorithm according to the training data acquired by the training data acquisition unit to acquire a trained anomaly identification model, including:
initializing model parameters of an anomaly identification model between a parameter upper limit and a parameter lower limit for a plurality of times to obtain a plurality of model parameter codes;
According to the training data acquired by the training data acquisition unit, acquiring the coding fitness corresponding to each model parameter code, and determining the mode parameter code with the maximum coding fitness as the current optimal code;
searching in the neighborhood of the current optimal code by adopting a first-level searching strategy to update the current optimal code, and obtaining the updated current optimal code;
based on the updated current optimal code, aiming at all model parameter codes, carrying out local search on the model parameter codes by adopting a secondary search strategy so as to acquire updated model parameter codes;
aiming at the updated model parameter codes, adopting a three-level search strategy to perform local and global balance search on the model parameter codes so as to obtain model parameter codes after secondary updating;
for the model parameter codes after the secondary updating, performing global searching on the model parameter codes by adopting a four-level searching strategy to obtain the model parameter codes after the tertiary updating;
judging whether the iteration ending condition is met or not based on the model parameter codes updated for three times, if so, ending the search, outputting the model parameter code with the minimum coding adaptability as the final parameter of the abnormal recognition model, obtaining the trained abnormal recognition model, otherwise, returning to the step of obtaining the current optimal code;
The search range level corresponding to the first-level search strategy to the fourth-level search strategy is increased progressively, so that the comprehensive search of the solution space is realized, and the globally optimal model parameter code is obtained.
In one possible implementation manner, a first-level search strategy is adopted to search in a neighborhood of a current optimal code so as to update the current optimal code, and the updated current optimal code is obtained, which comprises the following steps:
determining a first sub-optimal code and a second sub-optimal code, wherein the coding fitness of the first sub-optimal code is only smaller than the coding fitness of the current optimal code, and the coding fitness of the second sub-optimal code is only smaller than the coding fitness of the first sub-optimal code and the coding fitness of the current optimal code;
based on the current optimal code, the first suboptimal code and the second suboptimal code, the updated current optimal code is obtained by adopting information communication and a greedy algorithm.
In one possible implementation, based on the updated current optimal code, a local search is performed on the model parameter codes using a secondary search strategy for all the model parameter codes to obtain updated model parameter codes, including:
based on the updated current optimal code, searching is carried out by adopting a time memory mode based on the self position of the model parameter code aiming at all the model parameter codes, so that the optimal value of the model parameter code in a local range is determined, and the updated model parameter code is obtained.
In one possible implementation, for the updated model parameter codes, a three-level search strategy is used to perform local and global balance search on the model parameter codes to obtain a second updated model parameter code, including:
and aiming at the updated model parameter codes, performing guided search based on the current optimal codes and the average value of all the model parameter codes, so as to realize local and global balanced search and obtain the model parameter codes after secondary updating.
In one possible implementation, for the model parameter code after the second update, global searching is performed on the model parameter code by adopting a four-level searching strategy to obtain the model parameter code after the third update, including:
determining a plurality of inferior codes with minimum coding fitness aiming at the model parameter codes after the secondary updating; wherein the number of inferior codes is preset;
and performing global search on the inferior codes based on the Rhin flight and the spiral flight to obtain updated inferior codes, thereby obtaining model parameter codes after three updates.
The abnormality monitoring system of the power distribution automation machine room provided by the invention can realize the collection and intelligent analysis of multidimensional sensing data in the power distribution automation machine room, so that the intelligent monitoring of the machine room can be realized, the result of the intelligent analysis can be remotely shared to staff, the staff can timely grasp the parameters of the machine room operation, the abnormal condition of the machine room operation can be timely found, various abnormal conditions can be rapidly processed, and the property loss caused by the abnormal condition of the power distribution automation machine room can be effectively reduced.
Drawings
In order to more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the drawings that are needed in the examples will be briefly described below, it being understood that the following drawings only illustrate some examples of the present invention and therefore should not be considered as limiting the scope, and that other related drawings may be obtained from these drawings without inventive effort for a person skilled in the art. In the drawings:
fig. 1 is a schematic structural diagram of an anomaly monitoring system of a power distribution automation machine room according to an embodiment of the present invention.
The system comprises a 101-data sensing module, a 102-model acquisition module, a 103-data analysis module, a 104-online data storage module and a 105-remote query module.
Detailed Description
For the purpose of making apparent the objects, technical solutions and advantages of the present invention, the present invention will be further described in detail with reference to the following examples and the accompanying drawings, wherein the exemplary embodiments of the present invention and the descriptions thereof are for illustrating the present invention only and are not to be construed as limiting the present invention.
Example 1
As shown in fig. 1, an embodiment of the present invention provides an anomaly monitoring system for a power distribution automation machine room, which includes a data sensing module 101, a model acquisition module 102, a data analysis module 103, an online data storage module 104, and a remote query module 105.
The data sensing module 101 is configured to collect multidimensional sensing data of the power distribution automation machine room, and transmit the multidimensional sensing data to the data analysis module 103.
The data sensing module 101 may include a variety of sensors by which different sensed data in the power distribution automation room may be collected. For example: temperature sensors, voltage sensors, etc.
The model obtaining module 102 is configured to construct an anomaly recognition model by adopting multi-model association, train the anomaly recognition model by adopting a multi-stage reinforcement training algorithm, obtain a trained anomaly recognition model, and deploy the trained anomaly recognition model in the data analysis module 103.
The model acquisition module 102 may be software provided in the data processing apparatus, or may be a combination of software and hardware. The multi-model may refer to a plurality of neural network models, and an anomaly recognition model is constructed through the plurality of neural models, so that recognition of multi-dimensional sensing data can be realized. If the neural network model wants to have a better recognition effect, the neural network model needs to be trained first, and the final monitoring effect is determined by the quality of the training effect. In the prior art, the neural network model is often trained by adopting a gradient descent method and an Adam optimizer, and although the training speed is high, the training effect is poor, so that the final recognition effect of the neural network model is limited.
The data analysis module 103 is configured to schedule the pre-deployed trained anomaly identification model to analyze the multidimensional transmission data, obtain anomaly monitoring results of the power distribution automation and the power distribution automation method, and transmit the anomaly monitoring results to the online data storage module 104 for storage.
The data analysis module 103 may be software disposed in the data processing device, or may be a combination of software and hardware, and the data analysis module 103 has a capability of sending data, so that the data is stored online, and a worker can remotely check the machine room monitoring data and abnormal conditions.
The online data storage module 104 is configured to receive and store the abnormal monitoring result transmitted by the data analysis module 103, and share the abnormal monitoring result to the remote query module 105 in an online viewing manner.
Optionally, the abnormal monitoring result may include no abnormal condition and different abnormal categories, and when no abnormal condition exists, the abnormal monitoring result and the corresponding multidimensional sensing data are directly associated and stored, so that the online inspection by the staff is facilitated. When the abnormality monitoring result is that a certain number is common, alarm information is sent to the remote inquiry module 105 corresponding to the staff, so that the staff can rapidly process the abnormality, and larger loss is avoided.
The remote query module 105 is deployed on a mobile terminal of a worker, and is used for remotely accessing the abnormality monitoring result stored in the online data storage module 104.
Optionally, the remote query module 105 may schedule the mobile terminal of the staff to send out an audible alarm, a light alarm or a holy light alarm, so that the staff can know the abnormal situation in time.
The abnormality monitoring system of the power distribution automation machine room provided by the invention can realize the collection and intelligent analysis of multidimensional sensing data in the power distribution automation machine room, so that the intelligent monitoring of the machine room can be realized, the result of the intelligent analysis can be remotely shared to staff, the staff can timely grasp the parameters of the machine room operation, the abnormal condition of the machine room operation can be timely found, various abnormal conditions can be rapidly processed, and the property loss caused by the abnormal condition of the power distribution automation machine room can be effectively reduced.
In one possible implementation, the multi-dimensional sensing data includes a machine room temperature time sequence, a machine room power supply voltage time sequence, a water level sensing time sequence, a smoke sensing time sequence, and an image sensing time sequence, which are acquired at a preset sampling frequency.
Optionally, there may be a plurality of sensors of each type when acquiring the sensing data, so that the machine room temperature time sequence, the machine room power supply voltage time sequence, the water level sensing time sequence and the smoke sensing time sequence all contain the data of the plurality of sensors. For example: the sampling frequency is 5, the sampling period is 1 minute, the number of sensors is 3, and the time sequence of the machine room temperature acquired in 1 minute can be { A1, 40, 41, 39, 42, 41, A2, 42, 39, 41, A3, 40, 41, 39, 42, 41}, wherein A1, A2 and A3 are respectively unique numerical numbers corresponding to three different temperature sensors. It should be noted that the above sequences are only examples, and in practical applications, more sensors may be included and the sampling frequency may be set to be more or less.
In one possible implementation manner, the model acquisition module 102 includes a model generation unit, a training data acquisition unit, a model training unit, and a model transmission unit.
The model generation unit is used for generating sub-models respectively corresponding to the machine room temperature time sequence, the machine room power supply voltage time sequence, the water level sensing time sequence, the smoke sensing time sequence and the image sensing time sequence, and constructing an abnormal recognition model according to all the sub-models.
Optionally, the submodels corresponding to the machine room temperature time sequence, the machine room power supply voltage time sequence, the water level sensing time sequence and the smoke sensing time sequence respectively can be set as the BP neural network. The BP neural network is generally used for direct classification, and can input data in a vector form and output classification labels corresponding to sequences corresponding to different sequences. All data in the image sensing time sequence are images, so that the BP neural network is difficult to identify, and in the embodiment, each image in the image sensing time sequence is identified by adopting the convolutional neural network, so that the category corresponding to the image is determined.
The training data acquisition unit is used for acquiring training data input by man-machine interaction or pre-stored in a database.
The model training unit is used for training the abnormal recognition model by adopting a multi-stage reinforcement training algorithm according to the training data acquired by the training data acquisition unit to acquire the trained abnormal recognition model.
The model transmission unit is configured to deploy the trained anomaly identification model in the data analysis module 103.
In one possible implementation manner, before generating the sub-models respectively corresponding to the machine room temperature time sequence, the machine room power supply voltage time sequence, the water level sensing time sequence, the smoke sensing time sequence and the image sensing time sequence, and constructing the abnormality recognition model according to all the sub-models, the method further comprises:
And determining whether the temperature data exceeds a preset temperature threshold value according to each temperature data in the machine room temperature time sequence, if so, directly determining temperature abnormality to obtain a temperature abnormality monitoring result, and otherwise, constructing a submodel.
And determining whether the voltage data exceeds a preset voltage threshold value according to each voltage data in the machine room power supply voltage time sequence, if so, directly determining voltage abnormality to obtain a voltage abnormality monitoring result, and if not, constructing a submodel.
And determining whether the water level data exceeds a preset water level threshold value according to each water level data in the water level sensing time sequence, if so, directly determining water level abnormality to obtain a water level abnormality monitoring result, and otherwise, constructing a submodel.
And determining whether the smoke data exceeds a preset smoke threshold value according to each smoke data in the smoke sensing time sequence, if so, directly determining smoke abnormality to obtain a smoke abnormality monitoring result, and otherwise, constructing a submodel.
And determining a first abnormal monitoring result according to the temperature abnormal monitoring result, the voltage abnormal monitoring result, the water level abnormal monitoring result and the smoke abnormal monitoring result.
Since an anomaly in a parameter may be a false detection or a specific cause cannot be determined, the multidimensional parameter needs to be comprehensively identified to determine a final comprehensive anomaly class (for example, smoke is drawn by a person, and fire cannot be judged by smoke sensing data). However, when a certain parameter is abnormal, the warning device can also prompt the staff in time, thereby realizing the function of early warning.
In one possible implementation manner, generating sub-models respectively corresponding to the machine room temperature time sequence, the machine room power supply voltage time sequence, the water level sensing time sequence, the smoke sensing time sequence and the image sensing time sequence, and constructing an anomaly identification model according to all the sub-models, including:
generating a first feature extraction sub-model for extracting features in a machine room temperature time sequence, a machine room power supply voltage time sequence, a water level sensing time sequence and a smoke sensing time sequence, and generating a second feature extraction sub-model for extracting features in an image sensing time sequence; wherein each sequence corresponds to a first feature extraction sub-model.
And generating a feature recognition sub-model for carrying out fusion recognition on the features output by the first feature extraction sub-model and the second feature extraction sub-model.
And constructing an anomaly recognition model according to the first feature extraction sub-model, the second feature extraction sub-model and the feature recognition sub-model.
In this embodiment, the first feature extraction sub-model and the feature recognition sub-model may be set as a BP neural network, and the second feature extraction sub-model may be set as a convolutional neural network. The outputs of the first feature extraction submodel and the second feature extraction submodel form a vector, and the vector is used as the input of the feature recognition submodel, so that an abnormality recognition model can be obtained, and the output of the abnormality recognition model is the comprehensive abnormality type (namely, an abnormality monitoring result).
Therefore, it may be determined that the training data at least includes a historical machine room temperature time series, a machine room power supply voltage time series, a water level sensing time series, a smoke sensing time series, a temperature category corresponding to the machine room temperature time series (e.g., low temperature, normal temperature, high temperature, extremely high temperature, etc.), a voltage category corresponding to the machine room power supply voltage time series (e.g., no voltage, under voltage, normal, over voltage, etc.), a water level category corresponding to the water level sensing time series (e.g., no water, little water, a lot of water, etc.), a smoke category corresponding to the smoke sensing time series (e.g., no smoke, little smoke, a lot of smoke, etc.), an image sensing time series, an actual category corresponding to each image in the image sensing time series (e.g., no flame in an image, flame in an image), and a final comprehensive anomaly category (e.g., no anomaly, fire in a machine room, damage to a machine room device, fire in a machine room, a short circuit of water in a machine room, etc.).
When training, the non-image sequence is taken as input of the first feature extraction sub-model, and the category of the non-image sequence is taken as expected output of the first feature extraction sub-model; the image sequence serves as an input to the second feature extraction sub-model, and the class of the image sequence serves as a desired output of the second feature extraction sub-model.
It should be noted that the above-mentioned types are merely examples, and any fault associated with at least one sequence may be identified by using multidimensional sensing data, and the corresponding fault may be identified although some data is not useful. When the fault is associated with the multidimensional sensing data, the information implicit in the multidimensional sensing data can be fully mined, so that accurate fault identification is realized. Accordingly, various categories may be set according to actual demands, not limited to the example of the present embodiment.
When the machine room temperature time sequence, the machine room power supply voltage time sequence, the water level sensing time sequence and the smoke sensing time sequence can be identified once, and corresponding sequence types are obtained. The image sensing time sequence has a plurality of image data and cannot be identified at one time, so that a plurality of images can be identified one by adopting the second characteristic extraction sub-model to obtain a plurality of image categories, the sequence categories and the image categories form a vector, and the vector is used as the input of the characteristic identification sub-model, thereby obtaining an abnormal monitoring result.
In one possible implementation manner, training the anomaly identification model by using a multi-stage reinforcement training algorithm according to the training data acquired by the training data acquisition unit to acquire a trained anomaly identification model, including:
initializing model parameters of the anomaly identification model between the upper parameter limit and the lower parameter limit for a plurality of times, and obtaining a plurality of model parameter codes.
Optionally, since each dimension parameter has an upper parameter limit and a lower parameter limit, each time the code is updated, a parameter out-of-range process is required.
Although the data flow exists among the first feature extraction sub-model, the second feature extraction sub-model and the feature recognition sub-model, the training can be still carried out separately, and therefore the recognition effect of the data can be ensured. Therefore, this embodiment provides a better training manner, in which the model parameter codes for generating the first feature extraction sub-model are trained independently, the model parameter codes for generating the second feature extraction sub-model are trained independently, and the model parameter codes for generating the feature identification sub-model are trained independently, so that when training together is avoided, some model training is completed, but other model training is not completed, and finally, the training is slower, that is, the more parameters are, the slower the optimizing speed is.
In order to improve the training speed, in this embodiment, the first feature extraction sub-model, the second feature extraction sub-model, and the feature recognition sub-model may be trained in parallel, so that training time may be effectively saved.
And acquiring the coding fitness corresponding to each model parameter code according to the training data acquired by the training data acquisition unit, and determining the mode parameter code with the maximum coding fitness as the current optimal code.
Alternatively, when the first feature extraction sub-model and the feature recognition sub-model are separately trained, the sequence may be taken as input, the corresponding class may be taken as desired output, and the error function value of the sub-model may be obtained, while the second feature extraction sub-model is trained, the single image may be taken as input, the corresponding class may be taken as desired output, and the error function value of the sub-model may be obtained. And taking the inverse of the error function value as the coding fitness. The error function value is not normally zero, but in order to avoid zero denominator, the error function value may be added to a very small constant (e.g., 0.0001) and then inverted.
Alternatively, the error function value may be obtained using a root mean square error function or a cross entropy loss function, but other error functions may be used, just as examples.
And searching in the neighborhood of the current optimal code by adopting a first-level searching strategy to update the current optimal code and obtain the updated current optimal code.
Based on the updated current optimal code, a secondary search strategy is adopted for carrying out local search on the model parameter codes aiming at all the model parameter codes so as to acquire the updated model parameter codes.
And aiming at the updated model parameter codes, carrying out local and global balance search on the model parameter codes by adopting a three-level search strategy so as to obtain the model parameter codes after secondary updating.
And performing global search on the model parameter codes by adopting a four-level search strategy aiming at the model parameter codes after the secondary updating so as to obtain the model parameter codes after the tertiary updating.
And judging whether the iteration ending condition is met or not based on the model parameter codes updated for three times, if so, ending the search, outputting the model parameter code with the minimum coding adaptability as the final parameter of the abnormal recognition model, obtaining the trained abnormal recognition model, and otherwise, returning to the step of obtaining the current optimal code.
The search range level corresponding to the first-level search strategy to the fourth-level search strategy is increased progressively, so that the comprehensive search of the solution space is realized, and the globally optimal model parameter code is obtained.
The multi-stage reinforcement training algorithm provided by the embodiment can realize global search in a solution space so as to find the global optimal position, and compared with a gradient descent method, the multi-stage reinforcement training algorithm can effectively avoid sinking into local optimal, thereby realizing better training effect.
Alternatively, the iteration end condition may be a maximum training number or an error threshold, and when the training number is greater than or equal to the maximum training number or after the error function value is less than the error threshold, the model training may be considered complete.
In one possible implementation manner, a first-level search strategy is adopted to search in a neighborhood of a current optimal code so as to update the current optimal code, and the updated current optimal code is obtained, which comprises the following steps:
and determining a first sub-optimal code and a second sub-optimal code, wherein the coding fitness of the first sub-optimal code is only smaller than the coding fitness of the current optimal code, and the coding fitness of the second sub-optimal code is only smaller than the coding fitness of the first sub-optimal code and the coding fitness of the current optimal code.
Based on the current optimal code, the first suboptimal code and the second suboptimal code, the updated current optimal code is obtained by adopting information communication and a greedy algorithm.
Optionally, the present embodiment provides an example of a first-level search strategy, where the example includes:
the updated value of the current optimal code is determined as follows:
wherein X is best1 (t) represents the current optimal code during the t-th training process,updated value, w, representing the current optimal coding 1 X represents best1 Weight coefficient of (t), X best2 (t) represents the first suboptimal code in the t-th training process, w 2 X represents best2 Weight coefficient of (t), X best3 (t) represents a second suboptimal code, w 3 X represents best3 Weight coefficient of (t), u=1, 2, 3, f u Where u is 1, 2, 3 denote X respectively best1 (t)、X best2 (t)、X best3 (t) corresponding coding fitness.
Judging the updated value X b * est1 And (c) judging whether the coding fitness of (t) is greater than that of the original current optimal coding, if so, accepting the update, otherwise rejecting the update.
The method and the device not only improve information communication among codes, but also strengthen the influence degree of other codes on the global optimal position, quicken information communication in a population, search the optimal position to the greatest extent, and effectively reduce the possibility of early-maturing phenomenon of an iterative later-stage algorithm. In order to ensure the forward direction of the whole training process, the greedy algorithm is introduced to update, so that the training effect is prevented from being poor, and the training speed is ensured.
In one possible implementation, based on the updated current optimal code, a local search is performed on the model parameter codes using a secondary search strategy for all the model parameter codes to obtain updated model parameter codes, including:
based on the updated current optimal code, searching is carried out by adopting a time memory mode based on the self position of the model parameter code aiming at all the model parameter codes, so that the optimal value of the model parameter code in a local range is determined, and the updated model parameter code is obtained.
Optionally, the present embodiment provides an example of a secondary search strategy, where the example includes:
the update amount of the model parameter codes is determined as follows:
Δ i (t)＝X i (t)-X i (t-1)
wherein delta is i (t) represents the update amount of the ith model parameter code, X i (t) represents the ith model parameter code, X, in the t-th training process i (t-1) represents the ith model parameter code during the t-1 th training process;
according to the updating quantity of the model parameter code, acquiring updated model parameter codes as follows:
wherein X is i (t-2) represents the ith model parameter code, X in the t-2 training process i (t+1) represents updated X i (t)，X i (t-3) represents the ith model parameter code in the t-3 training process, and beta represents the adjustment coefficient.
When β is small, the previous activity of encoding will be ignored, and local extremum is easily trapped, but β is too large, the complexity of the algorithm will increase, and more time is spent. Thus, the adjustment coefficient β may be set as:
where e represents a natural constant and T represents the maximum number of training.
In the early stage of training, the two-stage searching strategy can promote the exploration capability of the algorithm by virtue of the memory characteristic, and the adjustment coefficient is continuously reduced along with the increase of the iteration times, so that the method is beneficial to the development of the later-stage algorithm. Therefore, the population diversity can be enhanced by combining a two-stage search strategy of adjusting the adjusting coefficient in a self-adaptive manner, the convergence speed of the algorithm can be improved, and the high-quality solution can be obtained.
In one possible implementation, for the updated model parameter codes, a three-level search strategy is used to perform local and global balance search on the model parameter codes to obtain a second updated model parameter code, including:
and aiming at the updated model parameter codes, performing guided search based on the current optimal codes and the average value of all the model parameter codes, so as to realize local and global balanced search and obtain the model parameter codes after secondary updating.
Optionally, the present embodiment provides an example of a three-level search strategy, where the example includes:
X i (t+1)'＝X i (t+1)*|sinR 1 |+R 2 *sinR 1 *|ξ 1 *X best (t+1)-ξ 2 *X avg (t+1)|
Wherein X is i (t+1)' represents X after the second update i (t+1)，R 1 Represents [0,2 pi ]]Random number between R 2 Represents [0, pi ]]Random number, X between best (t+1) represents the current optimum value, X after searching by the secondary search strategy avg (t+1) represents the mean value (i.e., each is the mean value) after the search of the secondary search strategy, ζ 1 ＝-π+(1-τ)*2π，ξ 1 Represents a first update coefficient, pi represents a circumference ratio, τ represents a golden section number, andξ 2 ＝-π+τ*2π，ξ 2 representing the second update coefficient.
After the random search stage, the improved golden sine idea is adopted to accelerate the optimizing speed of the algorithm, and the search range in the solution space is enlarged, so that the global search capability and convergence accuracy of the algorithm are improved.
In one possible implementation, for the model parameter code after the second update, global searching is performed on the model parameter code by adopting a four-level searching strategy to obtain the model parameter code after the third update, including:
determining a plurality of inferior codes with minimum coding fitness aiming at the model parameter codes after the secondary updating; wherein the number of inferior codes is preset;
and performing global search on the inferior codes based on the Rhin flight and the spiral flight to obtain updated inferior codes, thereby obtaining model parameter codes after three updates.
Optionally, the present embodiment provides an example of a four-level search strategy, where the example includes:
X worstε (t+1)＝X worst (t)+(X worst (t)-X levy )λlcos(2πl)+εA(t)
wherein X is worst (t) represents a bad code, X worstε (t+1) represents updated X worst (t)，X levy Represents the Rhin flight, lambda represents the upper limit lambda max And a lower limit value lambda min The random coefficient between, l represents [ -1,1 []The random number between them, ε represents [ -1,1 [ -1 ]]Random numbers in between, A (t) represents an adjustment step size, μ represents a first random flight coefficient, v represents a second random flight coefficient, η represents (0, 2)]Random numbers in between, and mu and v obey normal distribution, i.e. v-N (0, 1),σ μ represents intermediate parameters, and->Γ represents a gamma function.
The four-level search strategy provided by the embodiment searches at a certain rotation angle, avoids generating repeated individuals to the maximum extent, and can effectively improve the local development capability of an algorithm. The Rhin flight is a random walk mode, long jumps occasionally occur in the random walk process, and the Rhin flight can expand the group search range in the search process, and assist the algorithm to jump out of local optimum when necessary. The strategy not only can enhance the local development capability of the algorithm, but also can effectively assist the algorithm to jump out of local optimum and realize global search, thereby improving the searching performance of the algorithm.
It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
Those of ordinary skill in the art will appreciate that implementing all or part of the above facts and methods may be accomplished by a program to instruct related hardware, the program involved or the program may be stored in a computer readable storage medium, the program when executed comprising the steps of: the corresponding method steps are introduced at this time, and the storage medium may be a ROM/RAM, a magnetic disk, an optical disk, or the like.
The foregoing description of the embodiments has been provided for the purpose of illustrating the general principles of the invention, and is not meant to limit the scope of the invention, but to limit the invention to the particular embodiments, and any modifications, equivalents, improvements, etc. that fall within the spirit and principles of the invention are intended to be included within the scope of the invention.
Claims (10)
1. The abnormality monitoring system of the power distribution automation machine room is characterized by comprising a data sensing module, a model acquisition module, a data analysis module, an online data storage module and a remote inquiry module;
the data sensing module is used for collecting multidimensional sensing data of the power distribution automation machine room and transmitting the multidimensional sensing data to the data analysis module;
the model acquisition module is used for constructing an abnormal recognition model by adopting multi-model combination, training the abnormal recognition model by adopting a multi-stage reinforcement training algorithm, acquiring a trained abnormal recognition model, and deploying the trained abnormal recognition model in the data analysis module;
the data analysis module is used for scheduling a pre-deployed trained anomaly identification model to analyze multidimensional transmission data, acquiring anomaly monitoring results of power distribution automation and a method, and transmitting the anomaly monitoring results to the online data storage module for storage;
The online data storage module is used for receiving and storing the abnormal monitoring results transmitted by the data analysis module and sharing the abnormal monitoring results to the remote inquiry module in an online checking mode;
the remote inquiry module is deployed on a mobile terminal of a worker and is used for remotely accessing the abnormal monitoring result stored in the online data storage module.
2. The anomaly monitoring system of a power distribution automation machine room of claim 1, wherein the multi-dimensional sensing data comprises a machine room temperature time sequence, a machine room supply voltage time sequence, a water level sensing time sequence, a smoke sensing time sequence, and an image sensing time sequence acquired at a preset sampling frequency.
3. The anomaly monitoring system of the power distribution automation machine room of claim 2, wherein the model acquisition module comprises a model generation unit, a training data acquisition unit, a model training unit, and a model transmission unit;
the model generation unit is used for generating sub-models respectively corresponding to the machine room temperature time sequence, the machine room power supply voltage time sequence, the water level sensing time sequence, the smoke sensing time sequence and the image sensing time sequence, and constructing an abnormal recognition model according to all the sub-models;
The training data acquisition unit is used for acquiring training data input by man-machine interaction or pre-stored in a database;
the model training unit is used for training the abnormal recognition model by adopting a multi-stage reinforcement training algorithm according to the training data acquired by the training data acquisition unit to acquire a trained abnormal recognition model;
the model transmission unit is used for deploying the trained abnormal recognition model in the data analysis module.
4. The anomaly monitoring system of the power distribution automation machine room according to claim 3, wherein before generating sub-models respectively corresponding to the machine room temperature time sequence, the machine room power supply voltage time sequence, the water level sensing time sequence, the smoke sensing time sequence and the image sensing time sequence, and constructing the anomaly identification model according to all the sub-models, the anomaly monitoring system further comprises:
determining whether the temperature data exceeds a preset temperature threshold value according to each temperature data in the machine room temperature time sequence, if so, directly determining temperature abnormality to obtain a temperature abnormality monitoring result, otherwise, constructing a submodel;
for each voltage data in the machine room power supply voltage time sequence, determining whether the voltage data exceeds a preset voltage threshold, if so, directly determining voltage abnormality to obtain a voltage abnormality monitoring result, otherwise, constructing a sub-model;
For each water level data in the water level sensing time sequence, determining whether the water level data exceeds a preset water level threshold, if so, directly determining water level abnormality to obtain a water level abnormality monitoring result, otherwise, constructing a submodel;
for each smoke data in the smoke sensing time sequence, determining whether the smoke data exceeds a preset smoke threshold, if so, directly determining smoke abnormality to obtain a smoke abnormality monitoring result, otherwise, constructing a submodel;
and determining a first abnormal monitoring result according to the temperature abnormal monitoring result, the voltage abnormal monitoring result, the water level abnormal monitoring result and the smoke abnormal monitoring result.
5. The anomaly monitoring system of the power distribution automation machine room of claim 4, wherein generating sub-models respectively corresponding to the machine room temperature time sequence, the machine room power supply voltage time sequence, the water level sensing time sequence, the smoke sensing time sequence and the image sensing time sequence and constructing the anomaly identification model according to all the sub-models comprises:
generating a first feature extraction sub-model for extracting features in a machine room temperature time sequence, a machine room power supply voltage time sequence, a water level sensing time sequence and a smoke sensing time sequence, and generating a second feature extraction sub-model for extracting features in an image sensing time sequence;
Generating a feature recognition sub-model for performing fusion recognition on the features output by the first feature extraction sub-model and the second feature extraction sub-model;
and constructing an anomaly recognition model according to the first feature extraction sub-model, the second feature extraction sub-model and the feature recognition sub-model.
6. The abnormality monitoring system of a power distribution automation machine room according to claim 5, wherein training the abnormality recognition model by using a multi-stage reinforcement training algorithm based on the training data acquired by the training data acquisition unit, and acquiring the trained abnormality recognition model, comprises:
initializing model parameters of an anomaly identification model between a parameter upper limit and a parameter lower limit for a plurality of times to obtain a plurality of model parameter codes;
according to the training data acquired by the training data acquisition unit, acquiring the coding fitness corresponding to each model parameter code, and determining the mode parameter code with the maximum coding fitness as the current optimal code;
searching in the neighborhood of the current optimal code by adopting a first-level searching strategy to update the current optimal code, and obtaining the updated current optimal code;
based on the updated current optimal code, aiming at all model parameter codes, carrying out local search on the model parameter codes by adopting a secondary search strategy so as to acquire updated model parameter codes;
Aiming at the updated model parameter codes, adopting a three-level search strategy to perform local and global balance search on the model parameter codes so as to obtain model parameter codes after secondary updating;
for the model parameter codes after the secondary updating, performing global searching on the model parameter codes by adopting a four-level searching strategy to obtain the model parameter codes after the tertiary updating;
judging whether the iteration ending condition is met or not based on the model parameter codes updated for three times, if so, ending the search, outputting the model parameter code with the minimum coding adaptability as the final parameter of the abnormal recognition model, obtaining the trained abnormal recognition model, otherwise, returning to the step of obtaining the current optimal code;
the search range level corresponding to the first-level search strategy to the fourth-level search strategy is increased progressively, so that the comprehensive search of the solution space is realized, and the globally optimal model parameter code is obtained.
7. The anomaly monitoring system of the power distribution automation machine room of claim 6, wherein searching in a neighborhood of a current optimal code by using a first-level search strategy to update the current optimal code, and obtaining the updated current optimal code comprises:
Determining a first sub-optimal code and a second sub-optimal code, wherein the coding fitness of the first sub-optimal code is only smaller than the coding fitness of the current optimal code, and the coding fitness of the second sub-optimal code is only smaller than the coding fitness of the first sub-optimal code and the coding fitness of the current optimal code;
based on the current optimal code, the first suboptimal code and the second suboptimal code, the updated current optimal code is obtained by adopting information communication and a greedy algorithm.
8. The anomaly monitoring system of the power distribution automation machine room of claim 7, wherein the local search of the model parameter codes for all model parameter codes based on the updated current optimal code using a secondary search strategy to obtain updated model parameter codes comprises:
based on the updated current optimal code, searching is carried out by adopting a time memory mode based on the self position of the model parameter code aiming at all the model parameter codes, so that the optimal value of the model parameter code in a local range is determined, and the updated model parameter code is obtained.
9. The anomaly monitoring system of the power distribution automation machine room of claim 8, wherein the performing a local and global balance search on the model parameter codes using a three-level search strategy for updated model parameter codes to obtain a second updated model parameter code comprises:
And aiming at the updated model parameter codes, performing guided search based on the current optimal codes and the average value of all the model parameter codes, so as to realize local and global balanced search and obtain the model parameter codes after secondary updating.
10. The anomaly monitoring system of the power distribution automation machine room of claim 9, wherein for the model parameter codes after the secondary updating, global searching is performed on the model parameter codes by using a four-level search strategy to obtain the model parameter codes after the tertiary updating, comprising:
determining a plurality of inferior codes with minimum coding fitness aiming at the model parameter codes after the secondary updating; wherein the number of inferior codes is preset;
and performing global search on the inferior codes based on the Rhin flight and the spiral flight to obtain updated inferior codes, thereby obtaining model parameter codes after three updates.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311398096.7A CN117407817B (en) | 2023-10-25 | 2023-10-25 | An abnormal monitoring system for power distribution automation room |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311398096.7A CN117407817B (en) | 2023-10-25 | 2023-10-25 | An abnormal monitoring system for power distribution automation room |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117407817A true CN117407817A (en) | 2024-01-16 |
| CN117407817B CN117407817B (en) | 2024-07-09 |
Family
ID=89488616
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311398096.7A Active CN117407817B (en) | 2023-10-25 | 2023-10-25 | An abnormal monitoring system for power distribution automation room |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117407817B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117668671A (en) * | 2024-02-01 | 2024-03-08 | 成都工业学院 | An educational resource governance method based on machine learning |
| CN119597856A (en) * | 2024-11-14 | 2025-03-11 | 北京骏德卓越科技有限公司 | Geographic information data processing method |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109116833A (en) * | 2018-08-31 | 2019-01-01 | 重庆邮电大学 | Based on improvement drosophila-bat algorithm mechanical failure diagnostic method |
| CN110728075A (en) * | 2019-10-25 | 2020-01-24 | 山西应用科技学院 | Method for optimizing cryogenic process parameters by MOA algorithm |
| US20200387785A1 (en) * | 2019-06-05 | 2020-12-10 | Wuhan University | Power equipment fault detecting and positioning method of artificial intelligence inference fusion |
| CN114463594A (en) * | 2021-11-19 | 2022-05-10 | 中国华能集团清洁能源技术研究院有限公司 | Multi-mode deep learning power generation equipment abnormity integrated identification method and equipment |
| CN116610998A (en) * | 2023-05-24 | 2023-08-18 | 武汉恒达电气有限公司 | Switch cabinet fault diagnosis method and system based on multi-mode data fusion |
| CN116846085A (en) * | 2023-09-04 | 2023-10-03 | 山西绿柳科技有限公司 | A self-sensing monitoring intelligent distribution box control system |
-
2023
- 2023-10-25 CN CN202311398096.7A patent/CN117407817B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109116833A (en) * | 2018-08-31 | 2019-01-01 | 重庆邮电大学 | Based on improvement drosophila-bat algorithm mechanical failure diagnostic method |
| US20200387785A1 (en) * | 2019-06-05 | 2020-12-10 | Wuhan University | Power equipment fault detecting and positioning method of artificial intelligence inference fusion |
| CN110728075A (en) * | 2019-10-25 | 2020-01-24 | 山西应用科技学院 | Method for optimizing cryogenic process parameters by MOA algorithm |
| CN114463594A (en) * | 2021-11-19 | 2022-05-10 | 中国华能集团清洁能源技术研究院有限公司 | Multi-mode deep learning power generation equipment abnormity integrated identification method and equipment |
| CN116610998A (en) * | 2023-05-24 | 2023-08-18 | 武汉恒达电气有限公司 | Switch cabinet fault diagnosis method and system based on multi-mode data fusion |
| CN116846085A (en) * | 2023-09-04 | 2023-10-03 | 山西绿柳科技有限公司 | A self-sensing monitoring intelligent distribution box control system |
Non-Patent Citations (2)
| Title |
|---|
| CHARAF EDDINE KHAMOUDJ等: "A Learning Variable Neighborhood Search Approach for Induction Machines Bearing Failures Detection and Diagnosis", 《ENERGIES》, vol. 13, 9 June 2020 (2020-06-09), pages 1 - 30 * |
| 秦国领: "压缩域信号参数估计算法", 《电子信息对抗技术》, vol. 32, no. 5, 31 December 2017 (2017-12-31), pages 15 - 20 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117668671A (en) * | 2024-02-01 | 2024-03-08 | 成都工业学院 | An educational resource governance method based on machine learning |
| CN117668671B (en) * | 2024-02-01 | 2024-04-30 | 成都工业学院 | Educational resource management method based on machine learning |
| CN119597856A (en) * | 2024-11-14 | 2025-03-11 | 北京骏德卓越科技有限公司 | Geographic information data processing method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117407817B (en) | 2024-07-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN117407817B (en) | An abnormal monitoring system for power distribution automation room | |
| CN114879632B (en) | Multi-mode fusion vehicle fault diagnosis method and system based on big data | |
| CN110488150A (en) | An intelligent fault diagnosis method based on multi-algorithm fusion | |
| US20250341828A1 (en) | Methods and industrial internet of things systems for determining device anomaly based on anomaly threshold range | |
| CN113391239A (en) | Transformer abnormality monitoring method and system based on edge calculation | |
| CN117394529A (en) | Main distribution network loop-back reverse power supply control condition auxiliary decision-making method and system based on SCADA | |
| CN119760472A (en) | Intelligent anomaly detection and analysis method and system for electric power multi-mode space | |
| JP2017130094A (en) | Cell control device, and production system for managing operation situation of multiple manufacturing machines in manufacturing cell | |
| CN117725792A (en) | Building safety monitoring and early warning method based on time series and model modification | |
| CN116293454A (en) | A management system, method and medium based on dynamic balance of steam pipe network | |
| CN117923331B (en) | A load control system and method based on crane hoisting | |
| CN116866152A (en) | Risk operation management and control method and device, electronic equipment and storage medium | |
| CN120027365A (en) | Smart gas pipeline pressure difference safety monitoring Internet of Things system, method and storage medium | |
| CN120725344A (en) | Smart factory management methods, systems, equipment and media based on the Industrial Internet of Things | |
| CN120408465A (en) | Unmanned tower operation status monitoring system based on big data | |
| CN119222840B (en) | IoT-based remote monitoring and data acquisition system for cryogenic coupled heat pumps | |
| CN120450411A (en) | Methods, systems, equipment and media for risk identification and treatment of underground powerhouse construction | |
| CN117528551A (en) | Port operation network construction and data perception method and system | |
| CN118627884A (en) | Substation fault analysis and early warning method, device, storage medium and electronic equipment | |
| CN117922366A (en) | Electric vehicle charging pile remote monitoring method based on Internet of things | |
| CN118134462A (en) | Operation and maintenance method and device of power transformation equipment, electronic equipment and storage medium | |
| CN112737109B (en) | Real-time safety early warning system of smart power grids | |
| CN115310704A (en) | Multi-dimensional fire safety assessment early warning system based on system safety model | |
| CN119717548B (en) | A distributed control system and method for liquid hydrogen refueling station | |
| CN115095512B (en) | Tunnel air compressor unit intelligent control system and method based on artificial intelligence |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |